Generalized principal-agent problems, including Stackelberg games, contract design, and Bayesian persuasion, are a class of economic problems where an agent best responds to a principal's committed strategy. We study repeated generalized principal-agent problems under the assumption that the principal does not have commitment power and the agent uses algorithms to learn to respond to the principal. We reduce this problem to a one-shot generalized principal-agent problem with an approximately-best-responding agent. Using this reduction, we show that: (1) if the agent uses contextual no-regret learning algorithms, then the principal can guarantee a utility that is at least the principal's optimal utility in the classic non-learning model minus the square root of the agent's regret; (2) if the agent uses contextual no-swap-regret learning algorithms, then the principal cannot obtain any utility more than the optimal utility in the non-learning model plus the agent's swap regret. But (3) if the agent uses mean-based learning algorithms (which can be no-regret but not no-swap-regret), then the principal can do significantly better than the non-learning model. These general results not only refine previous results in Stackelberg games and contract design with learning agents but also lead to new results for Bayesian persuasion with a learning agent.
View on arXiv